I am trying to perform geometry optimization. I am using the coordinate that has been optimized using Gaussian as the input. However I am getting the following error.
ADIIS minimization failed. File a bug, and include your entire input and output !
! files.
Not sure, What is wrong. Kindly help me troubleshoot the problem. I am giving the input file below. Attached the output file in the first reply , filename = SixM_SN_3_output.dat.
Please let me know if you need any further information.
Input
import psi4
psi4.set_output_file(‘SixM_SN_3_output.dat’, True)
#! Sample m062x/6-31G** SixM_SN_3 Geom Opt
psi4.set_memory(‘50 GB’)
SixM_SN_3 = psi4.geometry(“”"
6 -1.424235000 -2.027927000 -0.690357000
6 -0.466164000 -1.016990000 -0.490713000
6 -0.875291000 0.338310000 -0.561876000
6 -2.221522000 0.623905000 -0.840286000
6 -3.096382000 -0.489634000 -1.035413000
6 0.888405000 -1.341236000 -0.210376000
6 0.061192000 1.386425000 -0.350442000
6 1.421149000 1.067621000 -0.072932000
6 1.834986000 -0.295980000 -0.004354000
6 2.353589000 2.104247000 0.135564000
6 1.900313000 3.453631000 0.063721000
6 0.591987000 3.755744000 -0.198973000
6 -0.369853000 2.726962000 -0.414298000
6 -1.745651000 2.995083000 -0.685590000
6 -2.643292000 1.982461000 -0.890483000
1 -3.688540000 2.201771000 -1.088666000
1 -2.072323000 4.030239000 -0.724928000
1 -4.147209000 -0.292990000 -1.250631000
1 2.620974000 4.250535000 0.224952000
1 0.268455000 4.791503000 -0.246984000
6 -1.007093000 -3.386952000 -0.608652000
6 0.297714000 -3.702838000 -0.343656000
6 1.282672000 -2.692078000 -0.137085000
1 -1.759033000 -4.152813000 -0.764728000
1 0.604919000 -4.743253000 -0.284980000
6 2.648583000 -2.984922000 0.142918000
6 3.560492000 -1.984483000 0.338961000
6 3.180721000 -0.612133000 0.272595000
1 2.957244000 -4.024911000 0.199694000
1 4.598130000 -2.225286000 0.552104000
6 3.708357000 1.759653000 0.411290000
1 4.426484000 2.558929000 0.570877000
6 4.105599000 0.452905000 0.476854000
1 5.142142000 0.206726000 0.689175000
7 -2.737089000 -1.745529000 -0.960960000
6 -3.349190000 -0.139882000 1.950364000
8 -4.312343000 0.473599000 1.724216000
8 -2.389423000 -0.750236000 2.189553000
“”")
psi4.optimize(‘m062x/6-31G**’, molecule=SixM_SN_3, maxiter=500)
scf_e, scf_wfn = psi4.frequency(‘scf/cc-pvdz’, molecule=SixM_SN_3, return_wfn=True)
fchk_writer = psi4.core.FCHKWriter(scf_wfn)
fchk_writer.write(‘SixM_SN_3.fchk’)
Thank you
So, the reason for the error is a known Psi4 bug: it doesn’t cause correctness errors, but it does make solving SCF during an optimization take far too long. It’s the same root issue behind some crashes we’ve seen. Psi4 is using orbital guess in a very foolish way. This can lead to problems early in an optimization. I’ll need to finally fix it. In the mean time, you should be able to bypass this with set scf_initial_accelerator none. Even then, it’ll be better when I finally fix the thing.
The other issue is that your input file is not sensible. For meaningful frequencies, you should almost always compute frequencies at the same level of theory as you use for geometry optimization. The one exception I know of is the recently published Concordant Mode Analysis, which is not what you’re doing here.
@jmisiewicz . Thank you for the kind reply. I am trying the job with the option scf_initial_accelerator set to None.
Thank you for pointing out the error in the input file. I have kept both geom opt and vib freq at the same level of theory. I hope the with the following commands, initially optimization is achieved and then the vibrational frequency is calculated on the optimized geometry. Please let me know if this is consistent. Thank you.
set scf scf_initial_accelerator none
psi4.optimize(‘m062x/6-31G**’, molecule=SixM_SN_3)
scf_e, scf_wfn = psi4.frequency(‘m062x/6-31G**’, molecule=SixM_SN_3, return_wfn=True)
fchk_writer = psi4.core.FCHKWriter(scf_wfn)
fchk_writer.write(‘SixM_SN_3.fchk’)
Yeah, I’d expect that to work. Let me know if it doesn’t.
Hopefully, I’ll get the root error fixed next week.
@jmisiewicz . The job failed saying “Could not converge SCF iterations in 100 iterations.”, even though I had given maxiter=500 for optimization. So this intrigued to look closer into the output fie. When I did, I found that the optimization job was converged, giving the following (SixM_SN_3_output.dat).
@DF-RKS iter 14: -1126.09326770452662 -2.36469e-11 6.24572e-09 DIIS
Energy and wave function converged.
Then I think, the frequency job started after that, however it errored out saying “Could not converge SCF iterations in 100 iterations.”
It looks like there is problem with the way the sequence of the job is given, initially optimization and then the frequency. I have attached both the input (6M_SN_3_input.txt) and output (SixM_SN_3_output.dat) files for reference.
Kindly suggest. Thank you.
SixM_SN_3_output.dat (76.2 KB)
6M_SN_3_input.txt (2.5 KB)
No, the frequency job did not start.
Replace set scf scf_initial_accelerator false with set scf guess_persist true. This new version of the code gets around the previously reported error in a more robust way.
@jmisiewicz . It throws the following error. Attached both i/p and o/p for reference.
SixM_SN_3_output.dat (2.3 KB)
→ core.set_local_option(“SCF”, “SCF_INITIAL_ACCELERATOR”, “false”)
core.set_local_option(“SCF”, “GUESS_PERSIST”, “true”)
psi4.optimize(‘m062x/6-31G**’, molecule=SixM_SN_3, maxiter=500)
scf_e, scf_wfn = psi4.frequency(‘m062x/6-31G**’, molecule=SixM_SN_3, return_wfn=True)
fchk_writer = psi4.core.FCHKWriter(scf_wfn)
fchk_writer.write(‘SixM_SN_3.fchk’)
!----------------------------------------------------------------------------------!
! !
! Fatal Error: FALSE is not a valid choice !
! Error occurred in file: /home/conda/feedstock_root/build_artifacts/psi4_16910215 !
! 40832/work/psi4/src/psi4/liboptions/liboptions.h on line: 50 !
! The most recent 5 function calls were: !
! psi::PsiException::PsiException(std::__cxx11::basic_string<char, !
! std::char_traits, std::allocator >, char const*, int) !
!
6M_SN_3.in.txt (2.5 KB)
No, remove anything to do with scf_initial_accelerator. Technically, I should have asked you to replace set scf_initial_accelerator none rather than set scf initial_accelerator false, but the point is the same: remove it.
I tried that as well. The input was as follows.
set scf_initial_accelerator none
set scf guess_persist true
psi4.optimize(‘m062x/6-31G**’, molecule=SixM_SN_3, maxiter=500)
scf_e, scf_wfn = psi4.frequency(‘m062x/6-31G**’, molecule=SixM_SN_3, return_wfn=True)
fchk_writer = psi4.core.FCHKWriter(scf_wfn)
fchk_writer.write(‘SixM_SN_3.fchk’)
However, got the following error. “Could not converge SCF iterations in 100 iterations.”
Attaching the files.
6M_SN_3.in.txt (2.5 KB)
SixM_SN_3_output.dat (75.2 KB)
…Okay, upon looking at this further, the optimizer is acting up and suggesting bad updates, so I’ll ping @AlexanderH. Internal coordinate steps can be horrible and lead to nonsense guess geometries. Cartesian steps take forever to converge and frequently make the energy more worse. Alex, I suggest you use set scf guess_persist true to bypass an (unrelated) issue with SCF guess reading across geometries.
I have no suggestions at this point.